Feature Extraction and Classification of Flaws in Radio Graphical Weld Images Using ANN

Abstract

In this paper, a novel approach for the detection and classification of flaws in weld images is presented. Computer based weld image analysis is most significant method. The method has been applied for detecting and discriminating flaws in the weld that may corresponds false alarms or all possible nine types of weld defects (Slag Inclusion, Wormhole, Porosity, Incomplete penetration, Under cuts, Cracks, Lack of fusion, Weaving fault Slag line), after being successfully tested on80 radiographic images obtained from EURECTEST, International scientific Association Brussels, Belgium, and 24 radiographs of ship weld provided by Technic Control Co. (Poland) were used, obtained from Ioannis Valavanis Greece.. The procedure to detect all the types of flaws and feature extraction is implemented by segmentation algorithm which can overcome computer complexity problem. Our problem focuses on the high performance classification by optimization of feature set by various selection algorithms like sequential forward search (SFS), sequential backward search algorithm (SBS) and sequential forward floating search algorithm (SFFS). Features are important for measuring parameters which leads in directional to understand image. We introduced 23 geometric features, and 14 texture features. The Experimental results show that our proposed method gives good performance of radiographic images.